Real-Time 3D Reconstruction of Multiple Guidewires at Dose Values of Conventional 2D Fluoroscopy

In interventional radiology, diagnostic or therapeutic procedures are performed minimally invasively using image guidance. Usually, a guidewire is introduced into the patient through a small incision or body orifice and navigated to the target site, followed by other interventional tools. Today, fluoroscopy, i.e. the display of a series of X-ray projections, can be considered a hallmark of image guidance. Providing sequences of two-dimensional images, it offers a high temporal resolution, but lacks depth information. For challenging procedures, 4D interventional guidance, i.e. the display of a series of 3D images, would be advantageous. The difficulty in realizing X-ray-based 4D interventional guidance lies in the development of a reconstruction algorithm that uses very little dose per 3D reconstruction, to enable image guidance with high temporal resolution during long interventions at acceptable accumulated dose. To this end, we here improve on a previously presented algorithm for the reconstruction of guidewires, stents and coils from four simultaneous X-ray projections. By incorporating temporal information into a 3D convolutional neural network, we reduce the number of X-ray projections that need to be acquired for the 3D reconstruction of guidewires from four to two, thereby halving dose and decreasing the demands put on imaging devices implementing the algorithm. In experiments on measured X-ray projections of two moving guidewires in an anthropomorphic phantom, we observe little deviation of our 3D reconstructions from the ground truth. To the best of our knowledge, this is the first algorithm capable of reconstructing multiple guidewires from only two update X-ray projections per 3D reconstruction.